Since April 2019, most federal government departments have been required to comply with the Directive on Automated Decision-Making. This policy instrument aims to reduce the risks associated with the deployment of automated decision systems used by the government to make administrative decisions concerning clients. All such systems undergo an algorithmic impact analysis, which assigns the system a risk score of 1 to 4. The directive requires that automated decision-making systems assigned a risk score of 2 or higher undergo a peer review.
The report “Peer Review for Automated Decision-Making Tools Under Canada's Directive on Automated Decision-Making”, written by faculty members Dr. Jason Millar and Dr. Kelly Bronson of the Centre for Law, Technology and Society, was a major reference in the development of the new TBS Guide to Peer Review of Automated Decision Systems, proposing improvements to the Directive on Automated Decision-Making.
The Guide adopts many of the CRAIEDL's recommendations to help the federal civil service comply with the Automated Decision-Making Directive when using AI and other automation tools.
CRAiEDL's involvement in this process ensures that AI is purchased, developed, and deployed responsibly by the Government of Canada, which has a lasting positive impact on the Canadian AI policy landscape. Since 2020, CRAiEDL has been working with the TBS and the Canada School of Public Service (CSPS) on the oversight of automated systems.
In May 2024, the CRAiEDL team, consisting of Dr. Jason Millar, Dr. Kelly Bronson, Dr. Jennie Day, Talia Cameron, Caitlin Heppner and Chantal Rodier, met as part of a participatory engagement process to discuss and review a draft of the Guide to the Peer Review of Automated Decision-Making Systems, providing essential feedback to the TBS for its development.
The Guide to Peer Review of Automated Decision-Making Systems is available on the Government of Canada website.